Apply customized function for extracting numbers from string to multiple columns in Python - python-3.x

Question

Given a dataset as follows:
id name value1 value2 value3
0 1 gz 1 6 st
1 2 sz 7-tower aka 278 3
2 3 wh NaN 67 s6.1
3 4 sh x3 45 34
I'd like to write a customized function to extract numbers from values columns.
Here is the pseudocode I have written:
def extract_nums(row):
return row.str.extract('(\d*\.?\d+)', expand=True)
df[['value1', 'value2', 'value3']] = df[['value1', 'value2', 'value3']].apply(extract_nums)
It raise an error:
ValueError: If using all scalar values, you must pass an index
Code for manipulating columns one by one works, but not concise:
df['value1'] = df['value1'].str.extract('(\d*\.?\d+)', expand=True)
df['value2'] = df['value2'].str.extract('(\d*\.?\d+)', expand=True)
df['value3'] = df['value3'].str.extract('(\d*\.?\d+)', expand=True)
How could write the code correctly? Thanks.

Answer 1

You can filter the value like columns then stack the columns and use str.extract to extract the numbers followed by unstack to reshape:
c = df.filter(like='value').columns
df[c] = df[c].stack().str.extract('(\d*\.?\d+)', expand=False).unstack()
Alternatively you can try str.extract with apply:
c = df.filter(like='value').columns
df[c] = df[c].apply(lambda s: s.str.extract('(\d*\.?\d+)', expand=False))
Result:
id name value1 value2 value3
0 1 gz 1 6 NaN
1 2 sz 7 278 3
2 3 wh NaN 67 6.1
3 4 sh 3 45 34

Answer 2

The following code also works out:
cols = ['value1', 'value2', 'value3']
for col in cols:
df[col] = df[col].str.extract('(\d*\.?\d+)', expand=True)
Alternative solution with function:
def extract_nums(row):
return row.str.extract('(\d*\.?\d+)', expand=False)
df[cols] = df[cols].apply(extract_nums)
Out:
id name value1 value2 value3
0 1 gz 1 6 NaN
1 2 sz 7 278 3
2 3 wh NaN 67 6.1
3 4 sh 3 45 34